Add nested models/ sources (src/models, tools/mtmd/models, vulkan-shaders models) missed by staging glob

Add build-info files, CI workflows, build scripts and chat templates missed by staging glob

README: this release is CUDA-only

Parakon runtime: llama.cpp fork with fused 1-bit CUDA/Metal/CPU kernels (part 9)

Parakon runtime: llama.cpp fork with fused 1-bit CUDA/Metal/CPU kernels (part 8)

Parakon runtime: llama.cpp fork with fused 1-bit CUDA/Metal/CPU kernels (part 7)

Parakon runtime: llama.cpp fork with fused 1-bit CUDA/Metal/CPU kernels (part 6)

Parakon runtime: llama.cpp fork with fused 1-bit CUDA/Metal/CPU kernels (part 5)

Parakon runtime: llama.cpp fork with fused 1-bit CUDA/Metal/CPU kernels (part 4)

Parakon runtime: llama.cpp fork with fused 1-bit CUDA/Metal/CPU kernels (part 3)

Parakon runtime: llama.cpp fork with fused 1-bit CUDA/Metal/CPU kernels (part 2)

Parakon runtime: llama.cpp fork with fused 1-bit CUDA/Metal/CPU kernels

initial commit